Video Describer

This project uses OpenAI's Vision API to describe YouTube videos.

Usage/Examples

curl --location 'http://localhost:8080/v1/video/describe' \
--header 'Content-Type: application/json' \
--data '{
    "videoID": "dQw4w9WgXcQ",
    "prompt": "These are frames of a video. Describe what happens in the video in one sentence, including a description of the environments where it takes place.",
    "end": "30s"
}'

Response:

The video features various scenes including a close-up of a person's shoe tapping, a man singing into a microphone with a graphic background, a smiling individual in a brick environment, a dancing woman with bright backlighting, a person posing in a denim outfit by a chain-link fence, a woman dancing in an urban outdoor setting, and the same man performing with different expressions in the brick environment and by the fence.

Environment Variables

Parameter	Default	Description
OPENAI_API_KEY	""	Required. API key from your OpenAI account.
SRV_HOST	0.0.0.0	Server host.
SRV_PORT	8080	Server port.
SRV_DEBUG	false	Enable debug logs.

Run Locally

Requirements:

• Go
• ffmpeg

Clone the project

  git clone https://github.com/WillRoque/video-describer

Go to the api directory

  cd cmd/api

Start the server

  go run .

API Reference

Describe a video

  POST /v1/video/describe

Parameter	Type	Description
videoID	string	Required. ID of a YouTube video.
prompt	string	Custom prompt to describe the video. The default one is "These are frames of a video. Describe what happens in the video in one sentence."
start	string	Describe the video starting from this position. Use the string representation of a time.Duration.
end	string	Describe the video up until this point. Use the string representation of a time.Duration.